We present an analysis of robotic grasping and apply it to the specific problem of using the Modular Dextrous Hand (MDH), developed at the Systems Research Center. The goal is to provide the capability to automatically select, analyze and implement a firm grasp for arbitrarily shaped objects. A geometrical grasp selection algorithm, based on the notion of form closure is presented and shown to have good properties. A method for analyzing grasps is presented and a method for determining finger torques and verifying grasps against task induced stability requirements is described. An algorithm to implement path planning to drive the hand to a grasp configuration while avoiding finger self-collision is presented. these capabilities are combined to provide a complete description of the grasping process. In addition, a description of the hardware and software architecture built to support the MDH is provided.